Can the model generated through LoRA training (using hv_train_network.py) be directly used for inference (hv_generate_video.py) without running merge_lora.py?

[JamesTensor]

Thank you very much for your world-class work!
If I want to continue LoRA training based on the already trained LoRA weights from the first stage, what should I do?
I've come up with this method: After merging the LoRA weights, obtain the merged DiT model, and then use hv_train_network.py to continue the LoRA training on the new DiT model.
Is this method correct? Or if there are other better methods, please let me know. Thank you very much!